For quite a while, people have been suggesting that AI would save the telcos. So far, that’s not happened, and in fact it’s far from clear how exactly such an outcome could come. Would AI add revenue? There’s no current evidence that AI generates more demand outside the data center. Would it reduce costs? So far, not a single telco has told me that AI or proposed AI projects had a significant impact on their costs. Improve capital efficiency? Claims but no proof.
The focus of telco AI has shifted to agents, which is at least a sensible transformation given enterprises’ own long-standing view that the generative AI model offered little promise for real transformation, but that what they considered an AI agent did have that potential. Great, but is there an indication that telcos have really followed the enterprise path? Sadly, not so far.
To start with, despite the fact that the telco OSS/BSS systems are very similar in both mission and features to the core applications of many verticals (particularly utilities), telcos have not seen AI as contributing to current workflows or augmenting existing applications. Enterprises, including utilities, saw this from the start, though the enterprise view of an AI agent wasn’t initially congruent with the technology specifics that have evolved. This gave enterprises a leg up; when the details of AI agents came along, they could fit them to a mission they were already seeing as valuable in business terms.
Second, unlike enterprises who watch across many verticals to find validation for new technologies, telcos tend to be very insular. Nobody has their issues, they believe, except other telcos. Even within a vertical, where enterprises unsurprisingly find the most convincing source of experience, telcos see only potential competitors. One telco planner, somewhat apologetic, said “Telcos think like utilities that don’t have any policy protection, because in a way we are just that.” This perhaps is why they fear competition more than value opportunity, or perhaps that feeling is why they are wary of other telcos.
There’s nothing wrong with introducing AI agents into telcos, any more than there is into any vertical. There’s no question that agents could improve telcos’ general business operations, make customer support cheaper and better, and make capital projects like 5G and 6G more efficient. All the telcos agree, but almost all (78 of 88 who have shared their views with me) agree that these changes won’t be transformational. Digging into that view, it seems to mean that there are long-term trends in the telcos’ business that create more negative forces than AI could offset.
Consider an example one telco type offered; the decision of SpaceX to buy $17 billion in spectrum to host cellular service, at least to expand direct-to-device services that today are largely confined to emergency backup for traditional cellular. LEO (low earth orbit) satellite is a viable competitor for traditional cell services, and many in the satellite telecom space believe that MEO (medium earth orbit) has even more potential. As voice calling gets less important (text and broadband Internet drive customer engagement), it’s very possible that satellite could grab an appreciable chunk of telco opportunity.
The worst impact here is that LEO/MEO satellite could be a great IoT connection approach, if we focus on the elements of IoT that aren’t exceptionally latency-sensitive. LEO/MEO has latencies that are far higher than those of terrestrial cellular connections, but for many applications that doesn’t matter because low-latency process control applications have long been hosted locally to the process being controlled. That’s not likely to change, so growth in IoT connection requirements would likely come only from some new application. As I noted earlier this week, telcos haven’t shown much interest in even identifying, much less deploying, useful tools in building out those new applications.
Given the fact that telcos are adverse to competition, this sure seems like a time to get smart. I get some feedback from telco planners that they are in fact seeing increased risk on the horizon, but the majority say that senior management is looking at 6G as the solution to their problems. The same people, in most cases, who looked to 5G for those answers already.
There’s another risk that telcos should be looking at, one that even the insightful planners aren’t talking about yet. It’s the micro/nano agent concept.
Telcos’ big opportunity in leveraging agent technology lies in their ability to find things to do with agents in their own access network, things that could justify their deployment. The relatively low ROI target that telcos typically apply to capital projects means they can invest in things that the OTT players, even most enterprises, can’t afford. If there is an opportunity to host agents proximate to the user, then having those kinds of agents out there already means telcos could have a head start on the opportunity. It’s hosting stuff off-premises that generates edge computing demand, and it’s edge computing that likely drives the next plausible generation of services. The problem with this happy opportunity for telcos is that micro/nano agents need only micro/nano hosting, which means that it’s very possible that the whole agent-hosting opportunity would be eaten by hosting them on premises.
Your phone, if you have a later-than-2023 model, is very likely to have at least some on-board AI capability. Companies are already talking about hosting AI agents on wearables. This opens a theoretical new model of AI, a model where way more could be done local to the application, meaning on the premises, on the user, in the vehicle. We already distribute computing, so we should expect to be able to distribute AI too, and while distributing anything involves communication, a distributed model will nearly always shortstop some traffic and reduce latency requirements for what remains. What this would do is convert a new revenue-generating role for telcos into a simple connectivity mission, that might be comfortable for them to contemplate but that would offer zero in terms of incremental profit potential.
There’s some good news here, though. Telcos like Deutsche Telecom have inked deals with Iridium to offer a direct satellite-to-IoT service. Given that telcos have always feared competition, and that another telco is the most credible competitor, this may start a wave of satellite integration into telco IoT plans. However, if this doesn’t go beyond simple connectivity, it still leaves telcos without a stepping-stone to exploiting AI agents in IoT, because the limiting factor in that advance is application validation, development, and hosting. This is an edge computing problem, making IoT and likely AI agents an edge computing opportunity for telcos. Creating their own agent applications is a potentially critical on-ramp to this capability, but one that telcos will have to take explicitly, and step out of competitive focus into that troubling-to-them opportunity zone.
